Your screen shows a lot: names, messages, balances, notifications, sometimes a password typed a little too slowly. Before you send any of it to an AI service, it's worth stopping to think. This guide covers AI privacy for screen content in plain words: what actually happens to your data, which questions to ask, and safer ways to get AI help.
What happens when you send something to a cloud AI
When an app uses a cloud AI, it sends your request — text, an image or audio — over the internet to the provider's servers. The model processes it and sends back an answer. Then, depending on the provider and your plan, the request may be:
- deleted quickly after processing,
- kept for a period for abuse monitoring,
- reviewed by staff in some cases,
- or used to improve future models, especially on some free tiers.
None of this is hidden; it's in each provider's terms. The problem is that few people read them before pasting in a screenshot.
How Cinematic Recorder handles your data
Here's what we can say for certain about our own app:
- Your videos stay on your phone. We never upload them to our servers.
- Your AI keys stay on your phone. They're stored in Android Keystore and never sent to our servers.
- AI is optional. With no key added, nothing goes to any AI provider.
- When AI runs, it goes direct. Requests go from your phone straight to the provider you chose, under that provider's terms.
Read the full details in our privacy policy. Note that captions are the one feature that sends anything out — the extracted sound, not the video — while privacy blur and the other AI touches are worked out on the phone itself. The AI editing assistant is still coming soon.
Questions to ask any AI provider
Before sending screen content to a provider, look for clear answers to these:
- Is my data used for training? Is there a setting to opt out? Is it different on free and paid plans?
- How long is it kept? Days, months, or not at all?
- Who can see it? Can staff review requests, and when?
- Where is it processed? This can matter for work or school data.
- Is there a middle layer? Services like OpenRouter pass your request to another host, so two sets of policies apply.
If you can't find answers, assume the less private option.
A simple traffic-light system for screen content
| Level | Examples | Send to cloud AI? |
|---|---|---|
| Green | Public app screens, your own tutorial script, a demo with fake data | Usually fine |
| Amber | Your name, business dashboards, client project names, private group chats | Only after blurring, or use local AI |
| Red | Passwords, OTPs, PINs, card or account numbers, ID documents, medical records | Never |
Safer choices that still give you AI help
1. Clean the recording first
Use a test account, turn on Do Not Disturb, and close private apps before recording. For a bKash or Nagad demo, use a test or demo account rather than your real wallet. Our checklist on privacy before you record covers this step by step.
2. Send text, not images
A transcript or your script reveals much less than a full screenshot. For titles and descriptions, text is all the AI needs.
3. Send only what's needed
If you need help with one section, send that section, not the whole video.
4. Use local AI for sensitive work
With Ollama running on your own computer, requests never reach an outside company; support for it in the app is coming soon. See private local AI with Ollama.
5. Choose paid tiers with clear terms
Paid API plans often have stronger privacy commitments than free ones. Check before you rely on it.
6. Or simply keep AI off
The recorder and editor work fully without AI. Auto zoom, cursor effects and reframing all run on your phone. See editing without AI.
Blurring and hiding information
Automatic privacy blur finds and covers sensitive areas for you, on the phone itself. Even so, the safest approach is to keep sensitive information off the screen while recording. You can also reframe or zoom in the editor so that private corners of the screen don't appear in the final export.
Remember that blur only protects the exported video. If you send the original, unblurred frames to an AI, the AI sees everything.
Recording other people's information
If your screen shows someone else's messages, contact details or photos, think about their privacy too. Ask permission when possible, or hide their details. This matters even more if you plan to send the recording to an AI service.
FAQ
Does Cinematic Recorder upload my videos to use with AI?
No. We never upload your videos. When an AI feature runs, it sends only what that feature needs, directly from your phone to the provider you chose.
Is local AI completely safe?
It keeps data away from AI companies, but your computer and network still need to be secure. Only open Ollama to a trusted home or office network.
Can AI providers see my API key?
Your provider uses the key to identify your account, as intended. Our servers never receive it; it stays in Android Keystore on your phone.
What should I never send to AI?
Passwords, OTPs, PINs, card and account numbers, ID documents and medical records. Keep them out of recordings too.
AI can be safe for screen content when you choose what to send. Keep red-level information off your screen, prefer text over images, and use local AI or AI off when in doubt.